// META: title=test WebNN API batchNormalization operation
// META: global=window,dedicatedworker
// META: variant=?cpu
// META: variant=?gpu
// META: variant=?npu
// META: script=../resources/utils.js
// META: timeout=long
'use strict';
// https://www.w3.org/TR/webnn/#api-mlgraphbuilder-batchnorm
// Normalize the values of the input tensor using Batch-Normalization.
//
// dictionary MLBatchNormalizationOptions {
// MLOperand scale;
// MLOperand bias;
// [EnforceRange] unsigned long axis = 1;
// double epsilon = 1e-5;
// };
//
// MLOperand batchNormalization(
// MLOperand input, MLOperand mean, MLOperand, variance,
// optional MLBatchNormalizationOptions options = {});
const getBatchNormPrecisionTolerance = (graphResources) => {
const toleranceValueDict = {float32: 6, float16: 6};
const expectedDataType =
getExpectedDataTypeOfSingleOutput(graphResources.expectedOutputs);
return {metricType: 'ULP', value: toleranceValueDict[expectedDataType]};
};
const batchNormTests = [
{
'name':
'batchNormalization float32 2D tensor (mean and variance are non-constant) default options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [4, 6], 'dataType': 'float32'}
},
'bnMean': {
'data': [
-7.814267635345459, -95.64129638671875, 38.15440368652344,
-55.95203399658203, -87.86500549316406, -41.63645553588867
],
'descriptor': {'dimensions': [6], 'dataType': 'float32'}
},
'bnVariance': {
'data': [
60.31186294555664, 26.43260383605957, 53.275634765625,
40.146121978759766, 59.41098403930664, 35.99981689453125
],
'descriptor': {'dimensions': [6], 'dataType': 'float32'}
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-4.312741756439209, 31.068212509155273, -13.910240173339844,
1.4459478855133057, 22.170541763305664, -6.407354354858398,
-6.995829105377197, 18.583200454711914, -10.831125259399414,
17.820920944213867, 16.2480411529541, 16.447195053100586,
11.57226848602295, 1.8526301383972168, 5.306026458740234,
24.145092010498047, 8.629376411437988, -9.216986656188965,
-0.1989477425813675, 34.203548431396484, -16.923160552978516,
18.671411514282227, 2.5159497261047363, 4.921559810638428
],
'descriptor': {'dimensions': [4, 6], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 2D constant tensor default options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [4, 6], 'dataType': 'float32'},
'constant': true
},
'bnMean': {
'data': [
-7.814267635345459, -95.64129638671875, 38.15440368652344,
-55.95203399658203, -87.86500549316406, -41.63645553588867
],
'descriptor': {'dimensions': [6], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [
60.31186294555664, 26.43260383605957, 53.275634765625,
40.146121978759766, 59.41098403930664, 35.99981689453125
],
'descriptor': {'dimensions': [6], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-4.312741756439209, 31.068212509155273, -13.910240173339844,
1.4459478855133057, 22.170541763305664, -6.407354354858398,
-6.995829105377197, 18.583200454711914, -10.831125259399414,
17.820920944213867, 16.2480411529541, 16.447195053100586,
11.57226848602295, 1.8526301383972168, 5.306026458740234,
24.145092010498047, 8.629376411437988, -9.216986656188965,
-0.1989477425813675, 34.203548431396484, -16.923160552978516,
18.671411514282227, 2.5159497261047363, 4.921559810638428
],
'descriptor': {'dimensions': [4, 6], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 2D tensor default options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [4, 6], 'dataType': 'float32'}
},
'bnMean': {
'data': [
-7.814267635345459, -95.64129638671875, 38.15440368652344,
-55.95203399658203, -87.86500549316406, -41.63645553588867
],
'descriptor': {'dimensions': [6], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [
60.31186294555664, 26.43260383605957, 53.275634765625,
40.146121978759766, 59.41098403930664, 35.99981689453125
],
'descriptor': {'dimensions': [6], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-4.312741756439209, 31.068212509155273, -13.910240173339844,
1.4459478855133057, 22.170541763305664, -6.407354354858398,
-6.995829105377197, 18.583200454711914, -10.831125259399414,
17.820920944213867, 16.2480411529541, 16.447195053100586,
11.57226848602295, 1.8526301383972168, 5.306026458740234,
24.145092010498047, 8.629376411437988, -9.216986656188965,
-0.1989477425813675, 34.203548431396484, -16.923160552978516,
18.671411514282227, 2.5159497261047363, 4.921559810638428
],
'descriptor': {'dimensions': [4, 6], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 3D tensor default options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [2, 3, 4], 'dataType': 'float32'}
},
'bnMean': {
'data': [12.810380935668945, 63.13715362548828, -61.62983322143555],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [18.358240127563477, 41.847232818603516, 16.12828254699707],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-12.630594253540039, 11.967890739440918, -17.781383514404297,
-13.910285949707031, 3.0739352703094482, -22.139259338378906,
-19.36661148071289, -9.775517463684082, 5.161267280578613,
29.53006935119629, 24.651947021484375, 29.550840377807617,
16.161500930786133, -23.088642120361328, 14.954023361206055,
19.656957626342773, -13.06058406829834, -24.745210647583008,
-11.206846237182617, 2.638929843902588, -5.910898208618164,
30.871898651123047, -1.7038332223892212, 12.331327438354492
],
'descriptor': {'dimensions': [2, 3, 4], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D tensor default options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-16.842504501342773, 2.2579827308654785, -20.842041015625,
-17.836172103881836, -1.7581257820129395, -19.30902862548828,
-17.37898826599121, -10.702629089355469, 6.4271392822265625,
17.812623977661133, 15.533489227294922, 17.822328567504883,
5.514280319213867, -24.963077545166016, 4.576685905456543,
8.228469848632812, -12.989363670349121, -21.123029708862305,
-11.698976516723633, -2.0609331130981445, 1.2540507316589355,
18.43954849243164, 3.2196571826934814, 9.777103424072266
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 5D tensor default options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [6, 1, 1, 2, 2], 'dataType': 'float32'}
},
'bnMean': {
'data': [35.4078254699707],
'descriptor': {'dimensions': [1], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [40.93109893798828],
'descriptor': {'dimensions': [1], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-11.990972518920898, 4.4829583168029785, -15.440524101257324,
-12.847999572753906, 7.442382335662842, -18.051416397094727,
-15.247910499572754, -5.550075531005859, -11.927642822265625,
3.369194269180298, 0.30708834528923035, 3.382232427597046,
7.291474342346191, -18.99486541748047, 6.4828104972839355,
9.632428169250488, -8.871702194213867, -20.686368942260742,
-6.99733304977417, 7.002535343170166, -18.877885818481445,
4.211489677429199, -16.237018585205078, -7.42683744430542
],
'descriptor': {'dimensions': [6, 1, 1, 2, 2], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D NCHW tensor options.axis=1',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'},
{'options': {'axis': 1}}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-16.842504501342773, 2.2579827308654785, -20.842041015625,
-17.836172103881836, -1.7581257820129395, -19.30902862548828,
-17.37898826599121, -10.702629089355469, 6.4271392822265625,
17.812623977661133, 15.533489227294922, 17.822328567504883,
5.514280319213867, -24.963077545166016, 4.576685905456543,
8.228469848632812, -12.989363670349121, -21.123029708862305,
-11.698976516723633, -2.0609331130981445, 1.2540507316589355,
18.43954849243164, 3.2196571826934814, 9.777103424072266
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D NHWC tensor options.axis=3',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 83.02227020263672, -40.90216064453125,
64.08863830566406, -80.08049011230469, 56.96306228637695,
-63.376670837402344, -62.144378662109375, 37.37249755859375,
-46.790367126464844, -0.10012771934270859, 57.046478271484375,
82.05680084228516, -21.35103988647461, -85.36802673339844,
-86.1164321899414, -96.93824005126953, 62.35185241699219,
76.8831787109375, -9.359310150146484, -68.4724349975586,
97.03362274169922, 80.20824432373047, -12.10716724395752
],
'descriptor': {'dimensions': [2, 2, 2, 3], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'},
{'options': {'axis': 3}}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-16.842504501342773, -1.7581257820129395, 6.4271392822265625,
2.2579827308654785, -19.30902862548828, 17.812623977661133,
-20.842041015625, -17.37898826599121, 15.533489227294922,
-17.836172103881836, -10.702629089355469, 17.822328567504883,
5.514280319213867, -12.989363670349121, 1.2540507316589355,
-24.963077545166016, -21.123029708862305, 18.43954849243164,
4.576685905456543, -11.698976516723633, 3.2196571826934814,
8.228469848632812, -2.0609331130981445, 9.777103424072266
],
'descriptor': {'dimensions': [2, 2, 2, 3], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D NCHW tensor options.scale',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnScale': {
'data': [65.50171661376953, -71.007568359375, -5.569730758666992],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'},
{'options': {'scale': 'bnScale'}}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-1103.212890625, 147.90174865722656, -1365.189453125,
-1168.2999267578125, 124.84024047851562, 1371.087158203125,
1234.0396728515625, 759.9676513671875, -35.79743576049805,
-99.2115249633789, -86.51734924316406, -99.26557159423828,
361.19482421875, -1635.1243896484375, 299.78076171875,
538.9788818359375, 922.3430786132812, 1499.89501953125,
830.7158813476562, 146.3418426513672, -6.984724998474121,
-102.70331573486328, -17.9326229095459, -54.455833435058594
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D NCHW tensor options.bias',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnBias': {
'data': [64.2044677734375, 75.28591918945312, -84.57243347167969],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'},
{'options': {'bias': 'bnBias'}}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
47.36196517944336, 66.46244812011719, 43.3624267578125,
46.36829376220703, 73.52779388427734, 55.976890563964844,
57.90693283081055, 64.58329010009766, -78.14529418945312,
-66.75981140136719, -69.03894805908203, -66.75010681152344,
69.71875, 39.241390228271484, 68.7811508178711,
72.43293762207031, 62.29655456542969, 54.16288757324219,
63.586944580078125, 73.22498321533203, -83.3183822631836,
-66.13288879394531, -81.35277557373047, -74.79533386230469
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D NCHW tensor options.epsilon',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 64.08863830566406, -63.376670837402344,
-46.790367126464844, 83.02227020263672, -80.08049011230469,
-62.144378662109375, -0.10012771934270859, -40.90216064453125,
56.96306228637695, 37.37249755859375, 57.046478271484375,
82.05680084228516, -86.1164321899414, 76.8831787109375,
97.03362274169922, -21.35103988647461, -96.93824005126953,
-9.359310150146484, 80.20824432373047, -85.36802673339844,
62.35185241699219, -68.4724349975586, -12.10716724395752
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'},
{'options': {'epsilon': 0.000001}}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-16.842506408691406, 2.2579832077026367, -20.842044830322266,
-17.8361759185791, -1.758125901222229, -19.309030532836914,
-17.37898826599121, -10.702629089355469, 6.427139759063721,
17.812625885009766, 15.533490180969238, 17.822330474853516,
5.514281272888184, -24.96308135986328, 4.576686382293701,
8.228470802307129, -12.989363670349121, -21.123031616210938,
-11.698976516723633, -2.0609331130981445, 1.254050850868225,
18.43954849243164, 3.2196574211120605, 9.777103424072266
],
'descriptor': {'dimensions': [2, 3, 2, 2], 'dataType': 'float32'}
}
}
}
},
{
'name': 'batchNormalization float32 4D NHWC tensor all options',
'graph': {
'inputs': {
'bnInput': {
'data': [
-41.30733108520508, 83.02227020263672, -40.90216064453125,
64.08863830566406, -80.08049011230469, 56.96306228637695,
-63.376670837402344, -62.144378662109375, 37.37249755859375,
-46.790367126464844, -0.10012771934270859, 57.046478271484375,
82.05680084228516, -21.35103988647461, -85.36802673339844,
-86.1164321899414, -96.93824005126953, 62.35185241699219,
76.8831787109375, -9.359310150146484, -68.4724349975586,
97.03362274169922, 80.20824432373047, -12.10716724395752
],
'descriptor': {'dimensions': [2, 2, 2, 3], 'dataType': 'float32'}
},
'bnMean': {
'data': [51.629150390625, 99.36075592041016, -96.1473617553711],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnVariance': {
'data': [30.448015213012695, 86.36219024658203, 73.88455200195312],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnScale': {
'data': [65.50171661376953, -71.007568359375, -5.569730758666992],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
},
'bnBias': {
'data': [64.2044677734375, 75.28591918945312, -84.57243347167969],
'descriptor': {'dimensions': [3], 'dataType': 'float32'},
'constant': true
}
},
'operators': [{
'name': 'batchNormalization',
'arguments': [
{'input': 'bnInput'}, {'mean': 'bnMean'}, {'variance': 'bnVariance'},
{
'options': {
'scale': 'bnScale',
'bias': 'bnBias',
'axis': 3,
'epsilon': 0.000001
}
}
],
'outputs': 'bnOutput'
}],
'expectedOutputs': {
'bnOutput': {
'data': [
-1039.0085734071204, 200.12613597546277, -120.36987167541395,
212.10626540432202, 1446.3732126569944, -183.78396479879416,
-1300.9852072279227, 1309.3257094058545, -171.08979404258523,
-1104.0956031373803, 835.2536189871761, -183.83801576309426,
425.3993215144054, 997.6290832897452, -91.55716013805052,
-1570.920072497096, 1575.1810627320297, -187.2757593197739,
363.98524710447384, 906.0018322105, -102.5050592863526,
603.1834043179756, 221.6277675074517, -139.02827100419768
],
'descriptor': {'dimensions': [2, 2, 2, 3], 'dataType': 'float32'}
}
}
}
}
];
if (navigator.ml) {
batchNormTests.forEach((test) => {
webnn_conformance_test(
buildGraphAndCompute, getBatchNormPrecisionTolerance, test);
});
} else {
test(() => assert_implements(navigator.ml, 'missing navigator.ml'));
}
Codebase Browser
chromium